qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
* [PULL 00/11] tcg patch queue
@ 2020-10-08 12:56 Richard Henderson
  2020-10-08 12:56 ` [PULL 01/11] tcg: Adjust simd_desc size encoding Richard Henderson
                   ` (12 more replies)
  0 siblings, 13 replies; 14+ messages in thread
From: Richard Henderson @ 2020-10-08 12:56 UTC (permalink / raw)
  To: qemu-devel; +Cc: peter.maydell

The following changes since commit 6eeea6725a70e6fcb5abba0764496bdab07ddfb3:

  Merge remote-tracking branch 'remotes/huth-gitlab/tags/pull-request-2020-10-06' into staging (2020-10-06 21:13:34 +0100)

are available in the Git repository at:

  https://github.com/rth7680/qemu.git tags/pull-tcg-20201008

for you to fetch changes up to 62475e9d007d83db4d0a6ccebcda8914f392e9c9:

  accel/tcg: Fix computing of is_write for MIPS (2020-10-08 05:57:32 -0500)

----------------------------------------------------------------
Extend maximum gvec vector size
Fix i386 avx2 dupi
Fix mips host user-only write detection
Misc cleanups.

----------------------------------------------------------------
Kele Huang (1):
      accel/tcg: Fix computing of is_write for MIPS

Richard Henderson (10):
      tcg: Adjust simd_desc size encoding
      tcg: Drop union from TCGArgConstraint
      tcg: Move sorted_args into TCGArgConstraint.sort_index
      tcg: Remove TCG_CT_REG
      tcg: Move some TCG_CT_* bits to TCGArgConstraint bitfields
      tcg: Remove TCGOpDef.used
      tcg/i386: Fix dupi for avx2 32-bit hosts
      tcg: Fix generation of dupi_vec for 32-bit host
      tcg/optimize: Fold dup2_vec
      tcg: Remove TCG_TARGET_HAS_cmp_vec

 include/tcg/tcg-gvec-desc.h  | 38 ++++++++++++------
 include/tcg/tcg.h            | 22 ++++------
 tcg/aarch64/tcg-target.h     |  1 -
 tcg/i386/tcg-target.h        |  1 -
 tcg/ppc/tcg-target.h         |  1 -
 accel/tcg/user-exec.c        | 43 ++++++++++++++++++--
 tcg/optimize.c               | 15 +++++++
 tcg/tcg-op-gvec.c            | 35 ++++++++++++----
 tcg/tcg-op-vec.c             | 12 ++++--
 tcg/tcg.c                    | 96 +++++++++++++++++++-------------------------
 tcg/aarch64/tcg-target.c.inc | 17 ++++----
 tcg/arm/tcg-target.c.inc     | 29 ++++++-------
 tcg/i386/tcg-target.c.inc    | 39 +++++++-----------
 tcg/mips/tcg-target.c.inc    | 21 +++++-----
 tcg/ppc/tcg-target.c.inc     | 29 ++++++-------
 tcg/riscv/tcg-target.c.inc   | 16 ++++----
 tcg/s390/tcg-target.c.inc    | 22 +++++-----
 tcg/sparc/tcg-target.c.inc   | 21 ++++------
 tcg/tci/tcg-target.c.inc     |  3 +-
 19 files changed, 244 insertions(+), 217 deletions(-)


^ permalink raw reply	[flat|nested] 14+ messages in thread

* [PULL 01/11] tcg: Adjust simd_desc size encoding
  2020-10-08 12:56 [PULL 00/11] tcg patch queue Richard Henderson
@ 2020-10-08 12:56 ` Richard Henderson
  2020-10-08 12:56 ` [PULL 02/11] tcg: Drop union from TCGArgConstraint Richard Henderson
                   ` (11 subsequent siblings)
  12 siblings, 0 replies; 14+ messages in thread
From: Richard Henderson @ 2020-10-08 12:56 UTC (permalink / raw)
  To: qemu-devel; +Cc: Frank Chang, peter.maydell, Philippe Mathieu-Daudé

With larger vector sizes, it turns out oprsz == maxsz, and we only
need to represent mismatch for oprsz <= 32.  We do, however, need
to represent larger oprsz and do so without reducing SIMD_DATA_BITS.

Reduce the size of the oprsz field and increase the maxsz field.
Steal the oprsz value of 24 to indicate equality with maxsz.

Tested-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/tcg/tcg-gvec-desc.h | 38 ++++++++++++++++++++++++-------------
 tcg/tcg-op-gvec.c           | 35 ++++++++++++++++++++++++++--------
 2 files changed, 52 insertions(+), 21 deletions(-)

diff --git a/include/tcg/tcg-gvec-desc.h b/include/tcg/tcg-gvec-desc.h
index 0224ac3e78..704bd86454 100644
--- a/include/tcg/tcg-gvec-desc.h
+++ b/include/tcg/tcg-gvec-desc.h
@@ -20,29 +20,41 @@
 #ifndef TCG_TCG_GVEC_DESC_H
 #define TCG_TCG_GVEC_DESC_H
 
-/* ??? These bit widths are set for ARM SVE, maxing out at 256 byte vectors. */
-#define SIMD_OPRSZ_SHIFT   0
-#define SIMD_OPRSZ_BITS    5
+/*
+ * This configuration allows MAXSZ to represent 2048 bytes, and
+ * OPRSZ to match MAXSZ, or represent the smaller values 8, 16, or 32.
+ *
+ * Encode this with:
+ *   0, 1, 3 -> 8, 16, 32
+ *   2       -> maxsz
+ *
+ * This steals the input that would otherwise map to 24 to match maxsz.
+ */
+#define SIMD_MAXSZ_SHIFT   0
+#define SIMD_MAXSZ_BITS    8
 
-#define SIMD_MAXSZ_SHIFT   (SIMD_OPRSZ_SHIFT + SIMD_OPRSZ_BITS)
-#define SIMD_MAXSZ_BITS    5
+#define SIMD_OPRSZ_SHIFT   (SIMD_MAXSZ_SHIFT + SIMD_MAXSZ_BITS)
+#define SIMD_OPRSZ_BITS    2
 
-#define SIMD_DATA_SHIFT    (SIMD_MAXSZ_SHIFT + SIMD_MAXSZ_BITS)
+#define SIMD_DATA_SHIFT    (SIMD_OPRSZ_SHIFT + SIMD_OPRSZ_BITS)
 #define SIMD_DATA_BITS     (32 - SIMD_DATA_SHIFT)
 
 /* Create a descriptor from components.  */
 uint32_t simd_desc(uint32_t oprsz, uint32_t maxsz, int32_t data);
 
-/* Extract the operation size from a descriptor.  */
-static inline intptr_t simd_oprsz(uint32_t desc)
-{
-    return (extract32(desc, SIMD_OPRSZ_SHIFT, SIMD_OPRSZ_BITS) + 1) * 8;
-}
-
 /* Extract the max vector size from a descriptor.  */
 static inline intptr_t simd_maxsz(uint32_t desc)
 {
-    return (extract32(desc, SIMD_MAXSZ_SHIFT, SIMD_MAXSZ_BITS) + 1) * 8;
+    return extract32(desc, SIMD_MAXSZ_SHIFT, SIMD_MAXSZ_BITS) * 8 + 8;
+}
+
+/* Extract the operation size from a descriptor.  */
+static inline intptr_t simd_oprsz(uint32_t desc)
+{
+    uint32_t f = extract32(desc, SIMD_OPRSZ_SHIFT, SIMD_OPRSZ_BITS);
+    intptr_t o = f * 8 + 8;
+    intptr_t m = simd_maxsz(desc);
+    return f == 2 ? m : o;
 }
 
 /* Extract the operation-specific data from a descriptor.  */
diff --git a/tcg/tcg-op-gvec.c b/tcg/tcg-op-gvec.c
index 7ebd9e8298..ddbe06b71a 100644
--- a/tcg/tcg-op-gvec.c
+++ b/tcg/tcg-op-gvec.c
@@ -37,11 +37,21 @@ static const TCGOpcode vecop_list_empty[1] = { 0 };
    of the operand offsets so that we can check them all at once.  */
 static void check_size_align(uint32_t oprsz, uint32_t maxsz, uint32_t ofs)
 {
-    uint32_t opr_align = oprsz >= 16 ? 15 : 7;
-    uint32_t max_align = maxsz >= 16 || oprsz >= 16 ? 15 : 7;
-    tcg_debug_assert(oprsz > 0);
-    tcg_debug_assert(oprsz <= maxsz);
-    tcg_debug_assert((oprsz & opr_align) == 0);
+    uint32_t max_align;
+
+    switch (oprsz) {
+    case 8:
+    case 16:
+    case 32:
+        tcg_debug_assert(oprsz <= maxsz);
+        break;
+    default:
+        tcg_debug_assert(oprsz == maxsz);
+        break;
+    }
+    tcg_debug_assert(maxsz <= (8 << SIMD_MAXSZ_BITS));
+
+    max_align = maxsz >= 16 ? 15 : 7;
     tcg_debug_assert((maxsz & max_align) == 0);
     tcg_debug_assert((ofs & max_align) == 0);
 }
@@ -77,12 +87,21 @@ uint32_t simd_desc(uint32_t oprsz, uint32_t maxsz, int32_t data)
 {
     uint32_t desc = 0;
 
-    assert(oprsz % 8 == 0 && oprsz <= (8 << SIMD_OPRSZ_BITS));
-    assert(maxsz % 8 == 0 && maxsz <= (8 << SIMD_MAXSZ_BITS));
-    assert(data == sextract32(data, 0, SIMD_DATA_BITS));
+    check_size_align(oprsz, maxsz, 0);
+    tcg_debug_assert(data == sextract32(data, 0, SIMD_DATA_BITS));
 
     oprsz = (oprsz / 8) - 1;
     maxsz = (maxsz / 8) - 1;
+
+    /*
+     * We have just asserted in check_size_align that either
+     * oprsz is {8,16,32} or matches maxsz.  Encode the final
+     * case with '2', as that would otherwise map to 24.
+     */
+    if (oprsz == maxsz) {
+        oprsz = 2;
+    }
+
     desc = deposit32(desc, SIMD_OPRSZ_SHIFT, SIMD_OPRSZ_BITS, oprsz);
     desc = deposit32(desc, SIMD_MAXSZ_SHIFT, SIMD_MAXSZ_BITS, maxsz);
     desc = deposit32(desc, SIMD_DATA_SHIFT, SIMD_DATA_BITS, data);
-- 
2.25.1



^ permalink raw reply related	[flat|nested] 14+ messages in thread

* [PULL 02/11] tcg: Drop union from TCGArgConstraint
  2020-10-08 12:56 [PULL 00/11] tcg patch queue Richard Henderson
  2020-10-08 12:56 ` [PULL 01/11] tcg: Adjust simd_desc size encoding Richard Henderson
@ 2020-10-08 12:56 ` Richard Henderson
  2020-10-08 12:56 ` [PULL 03/11] tcg: Move sorted_args into TCGArgConstraint.sort_index Richard Henderson
                   ` (10 subsequent siblings)
  12 siblings, 0 replies; 14+ messages in thread
From: Richard Henderson @ 2020-10-08 12:56 UTC (permalink / raw)
  To: qemu-devel; +Cc: peter.maydell, Philippe Mathieu-Daudé

The union is unused; let "regs" appear in the main structure
without the "u.regs" wrapping.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/tcg/tcg.h            |  4 +---
 tcg/tcg.c                    | 22 +++++++++++-----------
 tcg/aarch64/tcg-target.c.inc | 14 +++++++-------
 tcg/arm/tcg-target.c.inc     | 26 +++++++++++++-------------
 tcg/i386/tcg-target.c.inc    | 26 +++++++++++++-------------
 tcg/mips/tcg-target.c.inc    | 18 +++++++++---------
 tcg/ppc/tcg-target.c.inc     | 24 ++++++++++++------------
 tcg/riscv/tcg-target.c.inc   | 14 +++++++-------
 tcg/s390/tcg-target.c.inc    | 18 +++++++++---------
 tcg/sparc/tcg-target.c.inc   | 16 ++++++++--------
 tcg/tci/tcg-target.c.inc     |  2 +-
 11 files changed, 91 insertions(+), 93 deletions(-)

diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h
index 53ce94c2c5..a5a0ea4ada 100644
--- a/include/tcg/tcg.h
+++ b/include/tcg/tcg.h
@@ -985,9 +985,7 @@ void tcg_dump_op_count(void);
 typedef struct TCGArgConstraint {
     uint16_t ct;
     uint8_t alias_index;
-    union {
-        TCGRegSet regs;
-    } u;
+    TCGRegSet regs;
 } TCGArgConstraint;
 
 #define TCG_MAX_OP_ARGS 16
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 164a141d74..8e0df353ec 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -2210,7 +2210,7 @@ static int get_constraint_priority(const TCGOpDef *def, int k)
             return 0;
         n = 0;
         for(i = 0; i < TCG_TARGET_NB_REGS; i++) {
-            if (tcg_regset_test_reg(arg_ct->u.regs, i))
+            if (tcg_regset_test_reg(arg_ct->regs, i))
                 n++;
         }
     }
@@ -2268,7 +2268,7 @@ static void process_op_defs(TCGContext *s)
             /* Incomplete TCGTargetOpDef entry. */
             tcg_debug_assert(ct_str != NULL);
 
-            def->args_ct[i].u.regs = 0;
+            def->args_ct[i].regs = 0;
             def->args_ct[i].ct = 0;
             while (*ct_str != '\0') {
                 switch(*ct_str) {
@@ -2855,13 +2855,13 @@ static void liveness_pass_1(TCGContext *s)
                     pset = la_temp_pref(ts);
                     set = *pset;
 
-                    set &= ct->u.regs;
+                    set &= ct->regs;
                     if (ct->ct & TCG_CT_IALIAS) {
                         set &= op->output_pref[ct->alias_index];
                     }
                     /* If the combination is not possible, restart.  */
                     if (set == 0) {
-                        set = ct->u.regs;
+                        set = ct->regs;
                     }
                     *pset = set;
                 }
@@ -3551,8 +3551,8 @@ static void tcg_reg_alloc_dup(TCGContext *s, const TCGOp *op)
         return;
     }
 
-    dup_out_regs = tcg_op_defs[INDEX_op_dup_vec].args_ct[0].u.regs;
-    dup_in_regs = tcg_op_defs[INDEX_op_dup_vec].args_ct[1].u.regs;
+    dup_out_regs = tcg_op_defs[INDEX_op_dup_vec].args_ct[0].regs;
+    dup_in_regs = tcg_op_defs[INDEX_op_dup_vec].args_ct[1].regs;
 
     /* Allocate the output register now.  */
     if (ots->val_type != TEMP_VAL_REG) {
@@ -3706,10 +3706,10 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
             }
         }
 
-        temp_load(s, ts, arg_ct->u.regs, i_allocated_regs, i_preferred_regs);
+        temp_load(s, ts, arg_ct->regs, i_allocated_regs, i_preferred_regs);
         reg = ts->reg;
 
-        if (tcg_regset_test_reg(arg_ct->u.regs, reg)) {
+        if (tcg_regset_test_reg(arg_ct->regs, reg)) {
             /* nothing to do : the constraint is satisfied */
         } else {
         allocate_in_reg:
@@ -3717,7 +3717,7 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
                and move the temporary register into it */
             temp_load(s, ts, tcg_target_available_regs[ts->type],
                       i_allocated_regs, 0);
-            reg = tcg_reg_alloc(s, arg_ct->u.regs, i_allocated_regs,
+            reg = tcg_reg_alloc(s, arg_ct->regs, i_allocated_regs,
                                 o_preferred_regs, ts->indirect_base);
             if (!tcg_out_mov(s, ts->type, reg, ts->reg)) {
                 /*
@@ -3772,11 +3772,11 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
                 && !const_args[arg_ct->alias_index]) {
                 reg = new_args[arg_ct->alias_index];
             } else if (arg_ct->ct & TCG_CT_NEWREG) {
-                reg = tcg_reg_alloc(s, arg_ct->u.regs,
+                reg = tcg_reg_alloc(s, arg_ct->regs,
                                     i_allocated_regs | o_allocated_regs,
                                     op->output_pref[k], ts->indirect_base);
             } else {
-                reg = tcg_reg_alloc(s, arg_ct->u.regs, o_allocated_regs,
+                reg = tcg_reg_alloc(s, arg_ct->regs, o_allocated_regs,
                                     op->output_pref[k], ts->indirect_base);
             }
             tcg_regset_set_reg(o_allocated_regs, reg);
diff --git a/tcg/aarch64/tcg-target.c.inc b/tcg/aarch64/tcg-target.c.inc
index 2607fe4ab9..dbe5c6a14c 100644
--- a/tcg/aarch64/tcg-target.c.inc
+++ b/tcg/aarch64/tcg-target.c.inc
@@ -129,22 +129,22 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
     switch (*ct_str++) {
     case 'r': /* general registers */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs |= 0xffffffffu;
+        ct->regs |= 0xffffffffu;
         break;
     case 'w': /* advsimd registers */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs |= 0xffffffff00000000ull;
+        ct->regs |= 0xffffffff00000000ull;
         break;
     case 'l': /* qemu_ld / qemu_st address, data_reg */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffffffffu;
+        ct->regs = 0xffffffffu;
 #ifdef CONFIG_SOFTMMU
         /* x0 and x1 will be overwritten when reading the tlb entry,
            and x2, and x3 for helper args, better to avoid using them. */
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_X0);
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_X1);
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_X2);
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_X3);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_X0);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_X1);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_X2);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_X3);
 #endif
         break;
     case 'A': /* Valid for arithmetic immediate (positive or negative).  */
diff --git a/tcg/arm/tcg-target.c.inc b/tcg/arm/tcg-target.c.inc
index bc1e1b5a71..978eb1dd70 100644
--- a/tcg/arm/tcg-target.c.inc
+++ b/tcg/arm/tcg-target.c.inc
@@ -254,40 +254,40 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
 
     case 'r':
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffff;
+        ct->regs = 0xffff;
         break;
 
     /* qemu_ld address */
     case 'l':
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffff;
+        ct->regs = 0xffff;
 #ifdef CONFIG_SOFTMMU
         /* r0-r2,lr will be overwritten when reading the tlb entry,
            so don't use these. */
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R0);
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R1);
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R2);
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R3);
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R14);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R0);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R1);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R2);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R3);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R14);
 #endif
         break;
 
     /* qemu_st address & data */
     case 's':
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffff;
+        ct->regs = 0xffff;
         /* r0-r2 will be overwritten when reading the tlb entry (softmmu only)
            and r0-r1 doing the byte swapping, so don't use these. */
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R0);
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R1);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R0);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R1);
 #if defined(CONFIG_SOFTMMU)
         /* Avoid clashes with registers being used for helper args */
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R2);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R2);
 #if TARGET_LONG_BITS == 64
         /* Avoid clashes with registers being used for helper args */
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R3);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R3);
 #endif
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R14);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R14);
 #endif
         break;
 
diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc
index 0155c0691c..8661ec3393 100644
--- a/tcg/i386/tcg-target.c.inc
+++ b/tcg/i386/tcg-target.c.inc
@@ -209,42 +209,42 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
     switch(*ct_str++) {
     case 'a':
         ct->ct |= TCG_CT_REG;
-        tcg_regset_set_reg(ct->u.regs, TCG_REG_EAX);
+        tcg_regset_set_reg(ct->regs, TCG_REG_EAX);
         break;
     case 'b':
         ct->ct |= TCG_CT_REG;
-        tcg_regset_set_reg(ct->u.regs, TCG_REG_EBX);
+        tcg_regset_set_reg(ct->regs, TCG_REG_EBX);
         break;
     case 'c':
         ct->ct |= TCG_CT_REG;
-        tcg_regset_set_reg(ct->u.regs, TCG_REG_ECX);
+        tcg_regset_set_reg(ct->regs, TCG_REG_ECX);
         break;
     case 'd':
         ct->ct |= TCG_CT_REG;
-        tcg_regset_set_reg(ct->u.regs, TCG_REG_EDX);
+        tcg_regset_set_reg(ct->regs, TCG_REG_EDX);
         break;
     case 'S':
         ct->ct |= TCG_CT_REG;
-        tcg_regset_set_reg(ct->u.regs, TCG_REG_ESI);
+        tcg_regset_set_reg(ct->regs, TCG_REG_ESI);
         break;
     case 'D':
         ct->ct |= TCG_CT_REG;
-        tcg_regset_set_reg(ct->u.regs, TCG_REG_EDI);
+        tcg_regset_set_reg(ct->regs, TCG_REG_EDI);
         break;
     case 'q':
         /* A register that can be used as a byte operand.  */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = TCG_TARGET_REG_BITS == 64 ? 0xffff : 0xf;
+        ct->regs = TCG_TARGET_REG_BITS == 64 ? 0xffff : 0xf;
         break;
     case 'Q':
         /* A register with an addressable second byte (e.g. %ah).  */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xf;
+        ct->regs = 0xf;
         break;
     case 'r':
         /* A general register.  */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs |= ALL_GENERAL_REGS;
+        ct->regs |= ALL_GENERAL_REGS;
         break;
     case 'W':
         /* With TZCNT/LZCNT, we can have operand-size as an input.  */
@@ -253,15 +253,15 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
     case 'x':
         /* A vector register.  */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs |= ALL_VECTOR_REGS;
+        ct->regs |= ALL_VECTOR_REGS;
         break;
 
         /* qemu_ld/st address constraint */
     case 'L':
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = TCG_TARGET_REG_BITS == 64 ? 0xffff : 0xff;
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_L0);
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_L1);
+        ct->regs = TCG_TARGET_REG_BITS == 64 ? 0xffff : 0xff;
+        tcg_regset_reset_reg(ct->regs, TCG_REG_L0);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_L1);
         break;
 
     case 'e':
diff --git a/tcg/mips/tcg-target.c.inc b/tcg/mips/tcg-target.c.inc
index 7aa2073520..aae4fd187b 100644
--- a/tcg/mips/tcg-target.c.inc
+++ b/tcg/mips/tcg-target.c.inc
@@ -196,28 +196,28 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
     switch(*ct_str++) {
     case 'r':
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffffffff;
+        ct->regs = 0xffffffff;
         break;
     case 'L': /* qemu_ld input arg constraint */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffffffff;
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_A0);
+        ct->regs = 0xffffffff;
+        tcg_regset_reset_reg(ct->regs, TCG_REG_A0);
 #if defined(CONFIG_SOFTMMU)
         if (TCG_TARGET_REG_BITS < TARGET_LONG_BITS) {
-            tcg_regset_reset_reg(ct->u.regs, TCG_REG_A2);
+            tcg_regset_reset_reg(ct->regs, TCG_REG_A2);
         }
 #endif
         break;
     case 'S': /* qemu_st constraint */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffffffff;
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_A0);
+        ct->regs = 0xffffffff;
+        tcg_regset_reset_reg(ct->regs, TCG_REG_A0);
 #if defined(CONFIG_SOFTMMU)
         if (TCG_TARGET_REG_BITS < TARGET_LONG_BITS) {
-            tcg_regset_reset_reg(ct->u.regs, TCG_REG_A2);
-            tcg_regset_reset_reg(ct->u.regs, TCG_REG_A3);
+            tcg_regset_reset_reg(ct->regs, TCG_REG_A2);
+            tcg_regset_reset_reg(ct->regs, TCG_REG_A3);
         } else {
-            tcg_regset_reset_reg(ct->u.regs, TCG_REG_A1);
+            tcg_regset_reset_reg(ct->regs, TCG_REG_A1);
         }
 #endif
         break;
diff --git a/tcg/ppc/tcg-target.c.inc b/tcg/ppc/tcg-target.c.inc
index 7cb40b0466..0bd947b788 100644
--- a/tcg/ppc/tcg-target.c.inc
+++ b/tcg/ppc/tcg-target.c.inc
@@ -225,33 +225,33 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
     switch (*ct_str++) {
     case 'A': case 'B': case 'C': case 'D':
         ct->ct |= TCG_CT_REG;
-        tcg_regset_set_reg(ct->u.regs, 3 + ct_str[0] - 'A');
+        tcg_regset_set_reg(ct->regs, 3 + ct_str[0] - 'A');
         break;
     case 'r':
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffffffff;
+        ct->regs = 0xffffffff;
         break;
     case 'v':
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffffffff00000000ull;
+        ct->regs = 0xffffffff00000000ull;
         break;
     case 'L':                   /* qemu_ld constraint */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffffffff;
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R3);
+        ct->regs = 0xffffffff;
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R3);
 #ifdef CONFIG_SOFTMMU
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R4);
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R5);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R4);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R5);
 #endif
         break;
     case 'S':                   /* qemu_st constraint */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffffffff;
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R3);
+        ct->regs = 0xffffffff;
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R3);
 #ifdef CONFIG_SOFTMMU
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R4);
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R5);
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R6);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R4);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R5);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R6);
 #endif
         break;
     case 'I':
diff --git a/tcg/riscv/tcg-target.c.inc b/tcg/riscv/tcg-target.c.inc
index 2dfb07e247..0a69839adb 100644
--- a/tcg/riscv/tcg-target.c.inc
+++ b/tcg/riscv/tcg-target.c.inc
@@ -138,19 +138,19 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
     switch (*ct_str++) {
     case 'r':
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffffffff;
+        ct->regs = 0xffffffff;
         break;
     case 'L':
         /* qemu_ld/qemu_st constraint */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffffffff;
+        ct->regs = 0xffffffff;
         /* qemu_ld/qemu_st uses TCG_REG_TMP0 */
 #if defined(CONFIG_SOFTMMU)
-        tcg_regset_reset_reg(ct->u.regs, tcg_target_call_iarg_regs[0]);
-        tcg_regset_reset_reg(ct->u.regs, tcg_target_call_iarg_regs[1]);
-        tcg_regset_reset_reg(ct->u.regs, tcg_target_call_iarg_regs[2]);
-        tcg_regset_reset_reg(ct->u.regs, tcg_target_call_iarg_regs[3]);
-        tcg_regset_reset_reg(ct->u.regs, tcg_target_call_iarg_regs[4]);
+        tcg_regset_reset_reg(ct->regs, tcg_target_call_iarg_regs[0]);
+        tcg_regset_reset_reg(ct->regs, tcg_target_call_iarg_regs[1]);
+        tcg_regset_reset_reg(ct->regs, tcg_target_call_iarg_regs[2]);
+        tcg_regset_reset_reg(ct->regs, tcg_target_call_iarg_regs[3]);
+        tcg_regset_reset_reg(ct->regs, tcg_target_call_iarg_regs[4]);
 #endif
         break;
     case 'I':
diff --git a/tcg/s390/tcg-target.c.inc b/tcg/s390/tcg-target.c.inc
index 985115acfb..9cd266a2d0 100644
--- a/tcg/s390/tcg-target.c.inc
+++ b/tcg/s390/tcg-target.c.inc
@@ -409,24 +409,24 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
     switch (*ct_str++) {
     case 'r':                  /* all registers */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffff;
+        ct->regs = 0xffff;
         break;
     case 'L':                  /* qemu_ld/st constraint */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffff;
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R2);
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R3);
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_R4);
+        ct->regs = 0xffff;
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R2);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R3);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_R4);
         break;
     case 'a':                  /* force R2 for division */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0;
-        tcg_regset_set_reg(ct->u.regs, TCG_REG_R2);
+        ct->regs = 0;
+        tcg_regset_set_reg(ct->regs, TCG_REG_R2);
         break;
     case 'b':                  /* force R3 for division */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0;
-        tcg_regset_set_reg(ct->u.regs, TCG_REG_R3);
+        ct->regs = 0;
+        tcg_regset_set_reg(ct->regs, TCG_REG_R3);
         break;
     case 'A':
         ct->ct |= TCG_CT_CONST_S33;
diff --git a/tcg/sparc/tcg-target.c.inc b/tcg/sparc/tcg-target.c.inc
index 40bc12290c..e2de749af7 100644
--- a/tcg/sparc/tcg-target.c.inc
+++ b/tcg/sparc/tcg-target.c.inc
@@ -326,27 +326,27 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
     switch (*ct_str++) {
     case 'r':
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffffffff;
+        ct->regs = 0xffffffff;
         break;
     case 'R':
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = ALL_64;
+        ct->regs = ALL_64;
         break;
     case 'A': /* qemu_ld/st address constraint */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = TARGET_LONG_BITS == 64 ? ALL_64 : 0xffffffff;
+        ct->regs = TARGET_LONG_BITS == 64 ? ALL_64 : 0xffffffff;
     reserve_helpers:
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_O0);
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_O1);
-        tcg_regset_reset_reg(ct->u.regs, TCG_REG_O2);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_O0);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_O1);
+        tcg_regset_reset_reg(ct->regs, TCG_REG_O2);
         break;
     case 's': /* qemu_st data 32-bit constraint */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = 0xffffffff;
+        ct->regs = 0xffffffff;
         goto reserve_helpers;
     case 'S': /* qemu_st data 64-bit constraint */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = ALL_64;
+        ct->regs = ALL_64;
         goto reserve_helpers;
     case 'I':
         ct->ct |= TCG_CT_CONST_S11;
diff --git a/tcg/tci/tcg-target.c.inc b/tcg/tci/tcg-target.c.inc
index 992d50cb1e..a7215f346f 100644
--- a/tcg/tci/tcg-target.c.inc
+++ b/tcg/tci/tcg-target.c.inc
@@ -393,7 +393,7 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
     case 'L':                   /* qemu_ld constraint */
     case 'S':                   /* qemu_st constraint */
         ct->ct |= TCG_CT_REG;
-        ct->u.regs = BIT(TCG_TARGET_NB_REGS) - 1;
+        ct->regs = BIT(TCG_TARGET_NB_REGS) - 1;
         break;
     default:
         return NULL;
-- 
2.25.1



^ permalink raw reply related	[flat|nested] 14+ messages in thread

* [PULL 03/11] tcg: Move sorted_args into TCGArgConstraint.sort_index
  2020-10-08 12:56 [PULL 00/11] tcg patch queue Richard Henderson
  2020-10-08 12:56 ` [PULL 01/11] tcg: Adjust simd_desc size encoding Richard Henderson
  2020-10-08 12:56 ` [PULL 02/11] tcg: Drop union from TCGArgConstraint Richard Henderson
@ 2020-10-08 12:56 ` Richard Henderson
  2020-10-08 12:56 ` [PULL 04/11] tcg: Remove TCG_CT_REG Richard Henderson
                   ` (9 subsequent siblings)
  12 siblings, 0 replies; 14+ messages in thread
From: Richard Henderson @ 2020-10-08 12:56 UTC (permalink / raw)
  To: qemu-devel; +Cc: peter.maydell

This uses an existing hole in the TCGArgConstraint structure
and will be convenient for keeping the data in one place.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/tcg/tcg.h |  2 +-
 tcg/tcg.c         | 35 +++++++++++++++++------------------
 2 files changed, 18 insertions(+), 19 deletions(-)

diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h
index a5a0ea4ada..63955ac85b 100644
--- a/include/tcg/tcg.h
+++ b/include/tcg/tcg.h
@@ -985,6 +985,7 @@ void tcg_dump_op_count(void);
 typedef struct TCGArgConstraint {
     uint16_t ct;
     uint8_t alias_index;
+    uint8_t sort_index;
     TCGRegSet regs;
 } TCGArgConstraint;
 
@@ -1015,7 +1016,6 @@ typedef struct TCGOpDef {
     uint8_t nb_oargs, nb_iargs, nb_cargs, nb_args;
     uint8_t flags;
     TCGArgConstraint *args_ct;
-    int *sorted_args;
 #if defined(CONFIG_DEBUG_TCG)
     int used;
 #endif
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 8e0df353ec..da01e39fe4 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -944,7 +944,6 @@ void tcg_context_init(TCGContext *s)
     int op, total_args, n, i;
     TCGOpDef *def;
     TCGArgConstraint *args_ct;
-    int *sorted_args;
     TCGTemp *ts;
 
     memset(s, 0, sizeof(*s));
@@ -960,14 +959,11 @@ void tcg_context_init(TCGContext *s)
     }
 
     args_ct = g_malloc(sizeof(TCGArgConstraint) * total_args);
-    sorted_args = g_malloc(sizeof(int) * total_args);
 
     for(op = 0; op < NB_OPS; op++) {
         def = &tcg_op_defs[op];
         def->args_ct = args_ct;
-        def->sorted_args = sorted_args;
         n = def->nb_iargs + def->nb_oargs;
-        sorted_args += n;
         args_ct += n;
     }
 
@@ -2220,20 +2216,23 @@ static int get_constraint_priority(const TCGOpDef *def, int k)
 /* sort from highest priority to lowest */
 static void sort_constraints(TCGOpDef *def, int start, int n)
 {
-    int i, j, p1, p2, tmp;
+    int i, j;
+    TCGArgConstraint *a = def->args_ct;
 
-    for(i = 0; i < n; i++)
-        def->sorted_args[start + i] = start + i;
-    if (n <= 1)
+    for (i = 0; i < n; i++) {
+        a[start + i].sort_index = start + i;
+    }
+    if (n <= 1) {
         return;
-    for(i = 0; i < n - 1; i++) {
-        for(j = i + 1; j < n; j++) {
-            p1 = get_constraint_priority(def, def->sorted_args[start + i]);
-            p2 = get_constraint_priority(def, def->sorted_args[start + j]);
+    }
+    for (i = 0; i < n - 1; i++) {
+        for (j = i + 1; j < n; j++) {
+            int p1 = get_constraint_priority(def, a[start + i].sort_index);
+            int p2 = get_constraint_priority(def, a[start + j].sort_index);
             if (p1 < p2) {
-                tmp = def->sorted_args[start + i];
-                def->sorted_args[start + i] = def->sorted_args[start + j];
-                def->sorted_args[start + j] = tmp;
+                int tmp = a[start + i].sort_index;
+                a[start + i].sort_index = a[start + j].sort_index;
+                a[start + j].sort_index = tmp;
             }
         }
     }
@@ -3659,7 +3658,7 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
     for (k = 0; k < nb_iargs; k++) {
         TCGRegSet i_preferred_regs, o_preferred_regs;
 
-        i = def->sorted_args[nb_oargs + k];
+        i = def->args_ct[nb_oargs + k].sort_index;
         arg = op->args[i];
         arg_ct = &def->args_ct[i];
         ts = arg_temp(arg);
@@ -3695,7 +3694,7 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
                     int k2, i2;
                     reg = ts->reg;
                     for (k2 = 0 ; k2 < k ; k2++) {
-                        i2 = def->sorted_args[nb_oargs + k2];
+                        i2 = def->args_ct[nb_oargs + k2].sort_index;
                         if ((def->args_ct[i2].ct & TCG_CT_IALIAS) &&
                             reg == new_args[i2]) {
                             goto allocate_in_reg;
@@ -3760,7 +3759,7 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
         
         /* satisfy the output constraints */
         for(k = 0; k < nb_oargs; k++) {
-            i = def->sorted_args[k];
+            i = def->args_ct[k].sort_index;
             arg = op->args[i];
             arg_ct = &def->args_ct[i];
             ts = arg_temp(arg);
-- 
2.25.1



^ permalink raw reply related	[flat|nested] 14+ messages in thread

* [PULL 04/11] tcg: Remove TCG_CT_REG
  2020-10-08 12:56 [PULL 00/11] tcg patch queue Richard Henderson
                   ` (2 preceding siblings ...)
  2020-10-08 12:56 ` [PULL 03/11] tcg: Move sorted_args into TCGArgConstraint.sort_index Richard Henderson
@ 2020-10-08 12:56 ` Richard Henderson
  2020-10-08 12:56 ` [PULL 05/11] tcg: Move some TCG_CT_* bits to TCGArgConstraint bitfields Richard Henderson
                   ` (8 subsequent siblings)
  12 siblings, 0 replies; 14+ messages in thread
From: Richard Henderson @ 2020-10-08 12:56 UTC (permalink / raw)
  To: qemu-devel; +Cc: peter.maydell

This wasn't actually used for anything, really.  All variable
operands must accept registers, and which are indicated by the
set in TCGArgConstraint.regs.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/tcg/tcg.h            |  1 -
 tcg/tcg.c                    | 15 ++++-----------
 tcg/aarch64/tcg-target.c.inc |  3 ---
 tcg/arm/tcg-target.c.inc     |  3 ---
 tcg/i386/tcg-target.c.inc    | 11 -----------
 tcg/mips/tcg-target.c.inc    |  3 ---
 tcg/ppc/tcg-target.c.inc     |  5 -----
 tcg/riscv/tcg-target.c.inc   |  2 --
 tcg/s390/tcg-target.c.inc    |  4 ----
 tcg/sparc/tcg-target.c.inc   |  5 -----
 tcg/tci/tcg-target.c.inc     |  1 -
 11 files changed, 4 insertions(+), 49 deletions(-)

diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h
index 63955ac85b..3168315bb8 100644
--- a/include/tcg/tcg.h
+++ b/include/tcg/tcg.h
@@ -979,7 +979,6 @@ void tcg_dump_op_count(void);
 #define TCG_CT_ALIAS  0x80
 #define TCG_CT_IALIAS 0x40
 #define TCG_CT_NEWREG 0x20 /* output requires a new register */
-#define TCG_CT_REG    0x01
 #define TCG_CT_CONST  0x02 /* any constant of register size */
 
 typedef struct TCGArgConstraint {
diff --git a/tcg/tcg.c b/tcg/tcg.c
index da01e39fe4..55b2fc3ae3 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -2194,21 +2194,14 @@ static void tcg_dump_ops(TCGContext *s, bool have_prefs)
 /* we give more priority to constraints with less registers */
 static int get_constraint_priority(const TCGOpDef *def, int k)
 {
-    const TCGArgConstraint *arg_ct;
+    const TCGArgConstraint *arg_ct = &def->args_ct[k];
+    int n;
 
-    int i, n;
-    arg_ct = &def->args_ct[k];
     if (arg_ct->ct & TCG_CT_ALIAS) {
         /* an alias is equivalent to a single register */
         n = 1;
     } else {
-        if (!(arg_ct->ct & TCG_CT_REG))
-            return 0;
-        n = 0;
-        for(i = 0; i < TCG_TARGET_NB_REGS; i++) {
-            if (tcg_regset_test_reg(arg_ct->regs, i))
-                n++;
-        }
+        n = ctpop64(arg_ct->regs);
     }
     return TCG_TARGET_NB_REGS - n + 1;
 }
@@ -2276,7 +2269,7 @@ static void process_op_defs(TCGContext *s)
                         int oarg = *ct_str - '0';
                         tcg_debug_assert(ct_str == tdefs->args_ct_str[i]);
                         tcg_debug_assert(oarg < def->nb_oargs);
-                        tcg_debug_assert(def->args_ct[oarg].ct & TCG_CT_REG);
+                        tcg_debug_assert(def->args_ct[oarg].regs != 0);
                         /* TCG_CT_ALIAS is for the output arguments.
                            The input is tagged with TCG_CT_IALIAS. */
                         def->args_ct[i] = def->args_ct[oarg];
diff --git a/tcg/aarch64/tcg-target.c.inc b/tcg/aarch64/tcg-target.c.inc
index dbe5c6a14c..26f71cb599 100644
--- a/tcg/aarch64/tcg-target.c.inc
+++ b/tcg/aarch64/tcg-target.c.inc
@@ -128,15 +128,12 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
 {
     switch (*ct_str++) {
     case 'r': /* general registers */
-        ct->ct |= TCG_CT_REG;
         ct->regs |= 0xffffffffu;
         break;
     case 'w': /* advsimd registers */
-        ct->ct |= TCG_CT_REG;
         ct->regs |= 0xffffffff00000000ull;
         break;
     case 'l': /* qemu_ld / qemu_st address, data_reg */
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffffffffu;
 #ifdef CONFIG_SOFTMMU
         /* x0 and x1 will be overwritten when reading the tlb entry,
diff --git a/tcg/arm/tcg-target.c.inc b/tcg/arm/tcg-target.c.inc
index 978eb1dd70..62c37a954b 100644
--- a/tcg/arm/tcg-target.c.inc
+++ b/tcg/arm/tcg-target.c.inc
@@ -253,13 +253,11 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
         break;
 
     case 'r':
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffff;
         break;
 
     /* qemu_ld address */
     case 'l':
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffff;
 #ifdef CONFIG_SOFTMMU
         /* r0-r2,lr will be overwritten when reading the tlb entry,
@@ -274,7 +272,6 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
 
     /* qemu_st address & data */
     case 's':
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffff;
         /* r0-r2 will be overwritten when reading the tlb entry (softmmu only)
            and r0-r1 doing the byte swapping, so don't use these. */
diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc
index 8661ec3393..2f696074ab 100644
--- a/tcg/i386/tcg-target.c.inc
+++ b/tcg/i386/tcg-target.c.inc
@@ -208,42 +208,33 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
 {
     switch(*ct_str++) {
     case 'a':
-        ct->ct |= TCG_CT_REG;
         tcg_regset_set_reg(ct->regs, TCG_REG_EAX);
         break;
     case 'b':
-        ct->ct |= TCG_CT_REG;
         tcg_regset_set_reg(ct->regs, TCG_REG_EBX);
         break;
     case 'c':
-        ct->ct |= TCG_CT_REG;
         tcg_regset_set_reg(ct->regs, TCG_REG_ECX);
         break;
     case 'd':
-        ct->ct |= TCG_CT_REG;
         tcg_regset_set_reg(ct->regs, TCG_REG_EDX);
         break;
     case 'S':
-        ct->ct |= TCG_CT_REG;
         tcg_regset_set_reg(ct->regs, TCG_REG_ESI);
         break;
     case 'D':
-        ct->ct |= TCG_CT_REG;
         tcg_regset_set_reg(ct->regs, TCG_REG_EDI);
         break;
     case 'q':
         /* A register that can be used as a byte operand.  */
-        ct->ct |= TCG_CT_REG;
         ct->regs = TCG_TARGET_REG_BITS == 64 ? 0xffff : 0xf;
         break;
     case 'Q':
         /* A register with an addressable second byte (e.g. %ah).  */
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xf;
         break;
     case 'r':
         /* A general register.  */
-        ct->ct |= TCG_CT_REG;
         ct->regs |= ALL_GENERAL_REGS;
         break;
     case 'W':
@@ -252,13 +243,11 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
         break;
     case 'x':
         /* A vector register.  */
-        ct->ct |= TCG_CT_REG;
         ct->regs |= ALL_VECTOR_REGS;
         break;
 
         /* qemu_ld/st address constraint */
     case 'L':
-        ct->ct |= TCG_CT_REG;
         ct->regs = TCG_TARGET_REG_BITS == 64 ? 0xffff : 0xff;
         tcg_regset_reset_reg(ct->regs, TCG_REG_L0);
         tcg_regset_reset_reg(ct->regs, TCG_REG_L1);
diff --git a/tcg/mips/tcg-target.c.inc b/tcg/mips/tcg-target.c.inc
index aae4fd187b..41be574e89 100644
--- a/tcg/mips/tcg-target.c.inc
+++ b/tcg/mips/tcg-target.c.inc
@@ -195,11 +195,9 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
 {
     switch(*ct_str++) {
     case 'r':
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffffffff;
         break;
     case 'L': /* qemu_ld input arg constraint */
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffffffff;
         tcg_regset_reset_reg(ct->regs, TCG_REG_A0);
 #if defined(CONFIG_SOFTMMU)
@@ -209,7 +207,6 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
 #endif
         break;
     case 'S': /* qemu_st constraint */
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffffffff;
         tcg_regset_reset_reg(ct->regs, TCG_REG_A0);
 #if defined(CONFIG_SOFTMMU)
diff --git a/tcg/ppc/tcg-target.c.inc b/tcg/ppc/tcg-target.c.inc
index 0bd947b788..18ee989f95 100644
--- a/tcg/ppc/tcg-target.c.inc
+++ b/tcg/ppc/tcg-target.c.inc
@@ -224,19 +224,15 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
 {
     switch (*ct_str++) {
     case 'A': case 'B': case 'C': case 'D':
-        ct->ct |= TCG_CT_REG;
         tcg_regset_set_reg(ct->regs, 3 + ct_str[0] - 'A');
         break;
     case 'r':
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffffffff;
         break;
     case 'v':
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffffffff00000000ull;
         break;
     case 'L':                   /* qemu_ld constraint */
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffffffff;
         tcg_regset_reset_reg(ct->regs, TCG_REG_R3);
 #ifdef CONFIG_SOFTMMU
@@ -245,7 +241,6 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
 #endif
         break;
     case 'S':                   /* qemu_st constraint */
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffffffff;
         tcg_regset_reset_reg(ct->regs, TCG_REG_R3);
 #ifdef CONFIG_SOFTMMU
diff --git a/tcg/riscv/tcg-target.c.inc b/tcg/riscv/tcg-target.c.inc
index 0a69839adb..d536f3ccc1 100644
--- a/tcg/riscv/tcg-target.c.inc
+++ b/tcg/riscv/tcg-target.c.inc
@@ -137,12 +137,10 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
 {
     switch (*ct_str++) {
     case 'r':
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffffffff;
         break;
     case 'L':
         /* qemu_ld/qemu_st constraint */
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffffffff;
         /* qemu_ld/qemu_st uses TCG_REG_TMP0 */
 #if defined(CONFIG_SOFTMMU)
diff --git a/tcg/s390/tcg-target.c.inc b/tcg/s390/tcg-target.c.inc
index 9cd266a2d0..c5e096449b 100644
--- a/tcg/s390/tcg-target.c.inc
+++ b/tcg/s390/tcg-target.c.inc
@@ -408,23 +408,19 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
 {
     switch (*ct_str++) {
     case 'r':                  /* all registers */
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffff;
         break;
     case 'L':                  /* qemu_ld/st constraint */
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffff;
         tcg_regset_reset_reg(ct->regs, TCG_REG_R2);
         tcg_regset_reset_reg(ct->regs, TCG_REG_R3);
         tcg_regset_reset_reg(ct->regs, TCG_REG_R4);
         break;
     case 'a':                  /* force R2 for division */
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0;
         tcg_regset_set_reg(ct->regs, TCG_REG_R2);
         break;
     case 'b':                  /* force R3 for division */
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0;
         tcg_regset_set_reg(ct->regs, TCG_REG_R3);
         break;
diff --git a/tcg/sparc/tcg-target.c.inc b/tcg/sparc/tcg-target.c.inc
index e2de749af7..6775bd30fc 100644
--- a/tcg/sparc/tcg-target.c.inc
+++ b/tcg/sparc/tcg-target.c.inc
@@ -325,15 +325,12 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
 {
     switch (*ct_str++) {
     case 'r':
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffffffff;
         break;
     case 'R':
-        ct->ct |= TCG_CT_REG;
         ct->regs = ALL_64;
         break;
     case 'A': /* qemu_ld/st address constraint */
-        ct->ct |= TCG_CT_REG;
         ct->regs = TARGET_LONG_BITS == 64 ? ALL_64 : 0xffffffff;
     reserve_helpers:
         tcg_regset_reset_reg(ct->regs, TCG_REG_O0);
@@ -341,11 +338,9 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
         tcg_regset_reset_reg(ct->regs, TCG_REG_O2);
         break;
     case 's': /* qemu_st data 32-bit constraint */
-        ct->ct |= TCG_CT_REG;
         ct->regs = 0xffffffff;
         goto reserve_helpers;
     case 'S': /* qemu_st data 64-bit constraint */
-        ct->ct |= TCG_CT_REG;
         ct->regs = ALL_64;
         goto reserve_helpers;
     case 'I':
diff --git a/tcg/tci/tcg-target.c.inc b/tcg/tci/tcg-target.c.inc
index a7215f346f..231b9b1775 100644
--- a/tcg/tci/tcg-target.c.inc
+++ b/tcg/tci/tcg-target.c.inc
@@ -392,7 +392,6 @@ static const char *target_parse_constraint(TCGArgConstraint *ct,
     case 'r':
     case 'L':                   /* qemu_ld constraint */
     case 'S':                   /* qemu_st constraint */
-        ct->ct |= TCG_CT_REG;
         ct->regs = BIT(TCG_TARGET_NB_REGS) - 1;
         break;
     default:
-- 
2.25.1



^ permalink raw reply related	[flat|nested] 14+ messages in thread

* [PULL 05/11] tcg: Move some TCG_CT_* bits to TCGArgConstraint bitfields
  2020-10-08 12:56 [PULL 00/11] tcg patch queue Richard Henderson
                   ` (3 preceding siblings ...)
  2020-10-08 12:56 ` [PULL 04/11] tcg: Remove TCG_CT_REG Richard Henderson
@ 2020-10-08 12:56 ` Richard Henderson
  2020-10-08 12:56 ` [PULL 06/11] tcg: Remove TCGOpDef.used Richard Henderson
                   ` (7 subsequent siblings)
  12 siblings, 0 replies; 14+ messages in thread
From: Richard Henderson @ 2020-10-08 12:56 UTC (permalink / raw)
  To: qemu-devel; +Cc: peter.maydell

These are easier to set and test when they have their own fields.
Reduce the size of alias_index and sort_index to 4 bits, which is
sufficient for TCG_MAX_OP_ARGS.  This leaves only the bits indicating
constants within the ct field.

Move all initialization to allocation time, rather than init
individual fields in process_op_defs.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/tcg/tcg.h | 14 +++++++-------
 tcg/tcg.c         | 28 ++++++++++++----------------
 2 files changed, 19 insertions(+), 23 deletions(-)

diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h
index 3168315bb8..e8629b58c8 100644
--- a/include/tcg/tcg.h
+++ b/include/tcg/tcg.h
@@ -976,15 +976,15 @@ int64_t tcg_cpu_exec_time(void);
 void tcg_dump_info(void);
 void tcg_dump_op_count(void);
 
-#define TCG_CT_ALIAS  0x80
-#define TCG_CT_IALIAS 0x40
-#define TCG_CT_NEWREG 0x20 /* output requires a new register */
-#define TCG_CT_CONST  0x02 /* any constant of register size */
+#define TCG_CT_CONST  1 /* any constant of register size */
 
 typedef struct TCGArgConstraint {
-    uint16_t ct;
-    uint8_t alias_index;
-    uint8_t sort_index;
+    unsigned ct : 16;
+    unsigned alias_index : 4;
+    unsigned sort_index : 4;
+    bool oalias : 1;
+    bool ialias : 1;
+    bool newreg : 1;
     TCGRegSet regs;
 } TCGArgConstraint;
 
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 55b2fc3ae3..a8c28440e2 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -958,7 +958,7 @@ void tcg_context_init(TCGContext *s)
         total_args += n;
     }
 
-    args_ct = g_malloc(sizeof(TCGArgConstraint) * total_args);
+    args_ct = g_new0(TCGArgConstraint, total_args);
 
     for(op = 0; op < NB_OPS; op++) {
         def = &tcg_op_defs[op];
@@ -2197,7 +2197,7 @@ static int get_constraint_priority(const TCGOpDef *def, int k)
     const TCGArgConstraint *arg_ct = &def->args_ct[k];
     int n;
 
-    if (arg_ct->ct & TCG_CT_ALIAS) {
+    if (arg_ct->oalias) {
         /* an alias is equivalent to a single register */
         n = 1;
     } else {
@@ -2260,8 +2260,6 @@ static void process_op_defs(TCGContext *s)
             /* Incomplete TCGTargetOpDef entry. */
             tcg_debug_assert(ct_str != NULL);
 
-            def->args_ct[i].regs = 0;
-            def->args_ct[i].ct = 0;
             while (*ct_str != '\0') {
                 switch(*ct_str) {
                 case '0' ... '9':
@@ -2270,18 +2268,18 @@ static void process_op_defs(TCGContext *s)
                         tcg_debug_assert(ct_str == tdefs->args_ct_str[i]);
                         tcg_debug_assert(oarg < def->nb_oargs);
                         tcg_debug_assert(def->args_ct[oarg].regs != 0);
-                        /* TCG_CT_ALIAS is for the output arguments.
-                           The input is tagged with TCG_CT_IALIAS. */
                         def->args_ct[i] = def->args_ct[oarg];
-                        def->args_ct[oarg].ct |= TCG_CT_ALIAS;
+                        /* The output sets oalias.  */
+                        def->args_ct[oarg].oalias = true;
                         def->args_ct[oarg].alias_index = i;
-                        def->args_ct[i].ct |= TCG_CT_IALIAS;
+                        /* The input sets ialias. */
+                        def->args_ct[i].ialias = true;
                         def->args_ct[i].alias_index = oarg;
                     }
                     ct_str++;
                     break;
                 case '&':
-                    def->args_ct[i].ct |= TCG_CT_NEWREG;
+                    def->args_ct[i].newreg = true;
                     ct_str++;
                     break;
                 case 'i':
@@ -2848,7 +2846,7 @@ static void liveness_pass_1(TCGContext *s)
                     set = *pset;
 
                     set &= ct->regs;
-                    if (ct->ct & TCG_CT_IALIAS) {
+                    if (ct->ialias) {
                         set &= op->output_pref[ct->alias_index];
                     }
                     /* If the combination is not possible, restart.  */
@@ -3665,7 +3663,7 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
         }
 
         i_preferred_regs = o_preferred_regs = 0;
-        if (arg_ct->ct & TCG_CT_IALIAS) {
+        if (arg_ct->ialias) {
             o_preferred_regs = op->output_pref[arg_ct->alias_index];
             if (ts->fixed_reg) {
                 /* if fixed register, we must allocate a new register
@@ -3688,8 +3686,7 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
                     reg = ts->reg;
                     for (k2 = 0 ; k2 < k ; k2++) {
                         i2 = def->args_ct[nb_oargs + k2].sort_index;
-                        if ((def->args_ct[i2].ct & TCG_CT_IALIAS) &&
-                            reg == new_args[i2]) {
+                        if (def->args_ct[i2].ialias && reg == new_args[i2]) {
                             goto allocate_in_reg;
                         }
                     }
@@ -3760,10 +3757,9 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
             /* ENV should not be modified.  */
             tcg_debug_assert(!ts->fixed_reg);
 
-            if ((arg_ct->ct & TCG_CT_ALIAS)
-                && !const_args[arg_ct->alias_index]) {
+            if (arg_ct->oalias && !const_args[arg_ct->alias_index]) {
                 reg = new_args[arg_ct->alias_index];
-            } else if (arg_ct->ct & TCG_CT_NEWREG) {
+            } else if (arg_ct->newreg) {
                 reg = tcg_reg_alloc(s, arg_ct->regs,
                                     i_allocated_regs | o_allocated_regs,
                                     op->output_pref[k], ts->indirect_base);
-- 
2.25.1



^ permalink raw reply related	[flat|nested] 14+ messages in thread

* [PULL 06/11] tcg: Remove TCGOpDef.used
  2020-10-08 12:56 [PULL 00/11] tcg patch queue Richard Henderson
                   ` (4 preceding siblings ...)
  2020-10-08 12:56 ` [PULL 05/11] tcg: Move some TCG_CT_* bits to TCGArgConstraint bitfields Richard Henderson
@ 2020-10-08 12:56 ` Richard Henderson
  2020-10-08 12:56 ` [PULL 07/11] tcg/i386: Fix dupi for avx2 32-bit hosts Richard Henderson
                   ` (6 subsequent siblings)
  12 siblings, 0 replies; 14+ messages in thread
From: Richard Henderson @ 2020-10-08 12:56 UTC (permalink / raw)
  To: qemu-devel; +Cc: peter.maydell, Philippe Mathieu-Daudé

The last user of this field disappeared in f69d277ece4.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/tcg/tcg.h | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h
index e8629b58c8..8804a8c4a2 100644
--- a/include/tcg/tcg.h
+++ b/include/tcg/tcg.h
@@ -1015,9 +1015,6 @@ typedef struct TCGOpDef {
     uint8_t nb_oargs, nb_iargs, nb_cargs, nb_args;
     uint8_t flags;
     TCGArgConstraint *args_ct;
-#if defined(CONFIG_DEBUG_TCG)
-    int used;
-#endif
 } TCGOpDef;
 
 extern TCGOpDef tcg_op_defs[];
-- 
2.25.1



^ permalink raw reply related	[flat|nested] 14+ messages in thread

* [PULL 07/11] tcg/i386: Fix dupi for avx2 32-bit hosts
  2020-10-08 12:56 [PULL 00/11] tcg patch queue Richard Henderson
                   ` (5 preceding siblings ...)
  2020-10-08 12:56 ` [PULL 06/11] tcg: Remove TCGOpDef.used Richard Henderson
@ 2020-10-08 12:56 ` Richard Henderson
  2020-10-08 12:56 ` [PULL 08/11] tcg: Fix generation of dupi_vec for 32-bit host Richard Henderson
                   ` (5 subsequent siblings)
  12 siblings, 0 replies; 14+ messages in thread
From: Richard Henderson @ 2020-10-08 12:56 UTC (permalink / raw)
  To: qemu-devel; +Cc: peter.maydell, qemu-stable

The previous change wrongly stated that 32-bit avx2 should have
used VPBROADCASTW.  But that's a 16-bit broadcast and we want a
32-bit broadcast.

Fixes: 7b60ef3264e
Cc: qemu-stable@nongnu.org
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 tcg/i386/tcg-target.c.inc | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc
index 2f696074ab..d8797ed398 100644
--- a/tcg/i386/tcg-target.c.inc
+++ b/tcg/i386/tcg-target.c.inc
@@ -958,7 +958,7 @@ static void tcg_out_dupi_vec(TCGContext *s, TCGType type,
         new_pool_label(s, arg, R_386_PC32, s->code_ptr - 4, -4);
     } else {
         if (have_avx2) {
-            tcg_out_vex_modrm_pool(s, OPC_VPBROADCASTW + vex_l, ret);
+            tcg_out_vex_modrm_pool(s, OPC_VPBROADCASTD + vex_l, ret);
         } else {
             tcg_out_vex_modrm_pool(s, OPC_VBROADCASTSS, ret);
         }
-- 
2.25.1



^ permalink raw reply related	[flat|nested] 14+ messages in thread

* [PULL 08/11] tcg: Fix generation of dupi_vec for 32-bit host
  2020-10-08 12:56 [PULL 00/11] tcg patch queue Richard Henderson
                   ` (6 preceding siblings ...)
  2020-10-08 12:56 ` [PULL 07/11] tcg/i386: Fix dupi for avx2 32-bit hosts Richard Henderson
@ 2020-10-08 12:56 ` Richard Henderson
  2020-10-08 12:56 ` [PULL 09/11] tcg/optimize: Fold dup2_vec Richard Henderson
                   ` (4 subsequent siblings)
  12 siblings, 0 replies; 14+ messages in thread
From: Richard Henderson @ 2020-10-08 12:56 UTC (permalink / raw)
  To: qemu-devel; +Cc: peter.maydell, qemu-stable

The definition of INDEX_op_dupi_vec is that it operates on
units of tcg_target_ulong -- in this case 32 bits.  It does
not work to use this for a uint64_t value that happens to be
small enough to fit in tcg_target_ulong.

Fixes: d2fd745fe8b
Fixes: db432672dc5
Cc: qemu-stable@nongnu.org
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 tcg/tcg-op-vec.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)

diff --git a/tcg/tcg-op-vec.c b/tcg/tcg-op-vec.c
index ed6fb55fe1..cdbf11c573 100644
--- a/tcg/tcg-op-vec.c
+++ b/tcg/tcg-op-vec.c
@@ -252,10 +252,10 @@ TCGv_vec tcg_const_ones_vec_matching(TCGv_vec m)
 
 void tcg_gen_dup64i_vec(TCGv_vec r, uint64_t a)
 {
-    if (TCG_TARGET_REG_BITS == 32 && a == deposit64(a, 32, 32, a)) {
-        do_dupi_vec(r, MO_32, a);
-    } else if (TCG_TARGET_REG_BITS == 64 || a == (uint64_t)(int32_t)a) {
+    if (TCG_TARGET_REG_BITS == 64) {
         do_dupi_vec(r, MO_64, a);
+    } else if (a == dup_const(MO_32, a)) {
+        do_dupi_vec(r, MO_32, a);
     } else {
         TCGv_i64 c = tcg_const_i64(a);
         tcg_gen_dup_i64_vec(MO_64, r, c);
@@ -280,7 +280,11 @@ void tcg_gen_dup8i_vec(TCGv_vec r, uint32_t a)
 
 void tcg_gen_dupi_vec(unsigned vece, TCGv_vec r, uint64_t a)
 {
-    do_dupi_vec(r, MO_REG, dup_const(vece, a));
+    if (vece == MO_64) {
+        tcg_gen_dup64i_vec(r, a);
+    } else {
+        do_dupi_vec(r, MO_REG, dup_const(vece, a));
+    }
 }
 
 void tcg_gen_dup_i64_vec(unsigned vece, TCGv_vec r, TCGv_i64 a)
-- 
2.25.1



^ permalink raw reply related	[flat|nested] 14+ messages in thread

* [PULL 09/11] tcg/optimize: Fold dup2_vec
  2020-10-08 12:56 [PULL 00/11] tcg patch queue Richard Henderson
                   ` (7 preceding siblings ...)
  2020-10-08 12:56 ` [PULL 08/11] tcg: Fix generation of dupi_vec for 32-bit host Richard Henderson
@ 2020-10-08 12:56 ` Richard Henderson
  2020-10-08 12:56 ` [PULL 10/11] tcg: Remove TCG_TARGET_HAS_cmp_vec Richard Henderson
                   ` (3 subsequent siblings)
  12 siblings, 0 replies; 14+ messages in thread
From: Richard Henderson @ 2020-10-08 12:56 UTC (permalink / raw)
  To: qemu-devel; +Cc: peter.maydell

When the two arguments are identical, this can be reduced to
dup_vec or to mov_vec from a tcg_constant_vec.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 tcg/optimize.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/tcg/optimize.c b/tcg/optimize.c
index 53aa8e5329..220f4601d5 100644
--- a/tcg/optimize.c
+++ b/tcg/optimize.c
@@ -1109,6 +1109,21 @@ void tcg_optimize(TCGContext *s)
             }
             goto do_default;
 
+        case INDEX_op_dup2_vec:
+            assert(TCG_TARGET_REG_BITS == 32);
+            if (arg_is_const(op->args[1]) && arg_is_const(op->args[2])) {
+                tmp = arg_info(op->args[1])->val;
+                if (tmp == arg_info(op->args[2])->val) {
+                    tcg_opt_gen_movi(s, op, op->args[0], tmp);
+                    break;
+                }
+            } else if (args_are_copies(op->args[1], op->args[2])) {
+                op->opc = INDEX_op_dup_vec;
+                TCGOP_VECE(op) = MO_32;
+                nb_iargs = 1;
+            }
+            goto do_default;
+
         CASE_OP_32_64(not):
         CASE_OP_32_64(neg):
         CASE_OP_32_64(ext8s):
-- 
2.25.1



^ permalink raw reply related	[flat|nested] 14+ messages in thread

* [PULL 10/11] tcg: Remove TCG_TARGET_HAS_cmp_vec
  2020-10-08 12:56 [PULL 00/11] tcg patch queue Richard Henderson
                   ` (8 preceding siblings ...)
  2020-10-08 12:56 ` [PULL 09/11] tcg/optimize: Fold dup2_vec Richard Henderson
@ 2020-10-08 12:56 ` Richard Henderson
  2020-10-08 12:56 ` [PULL 11/11] accel/tcg: Fix computing of is_write for MIPS Richard Henderson
                   ` (2 subsequent siblings)
  12 siblings, 0 replies; 14+ messages in thread
From: Richard Henderson @ 2020-10-08 12:56 UTC (permalink / raw)
  To: qemu-devel; +Cc: peter.maydell, Philippe Mathieu-Daudé

The cmp_vec opcode is mandatory; this symbol is unused.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 tcg/aarch64/tcg-target.h | 1 -
 tcg/i386/tcg-target.h    | 1 -
 tcg/ppc/tcg-target.h     | 1 -
 3 files changed, 3 deletions(-)

diff --git a/tcg/aarch64/tcg-target.h b/tcg/aarch64/tcg-target.h
index 9bc2a5ecbe..663dd0b95e 100644
--- a/tcg/aarch64/tcg-target.h
+++ b/tcg/aarch64/tcg-target.h
@@ -139,7 +139,6 @@ typedef enum {
 #define TCG_TARGET_HAS_shi_vec          1
 #define TCG_TARGET_HAS_shs_vec          0
 #define TCG_TARGET_HAS_shv_vec          1
-#define TCG_TARGET_HAS_cmp_vec          1
 #define TCG_TARGET_HAS_mul_vec          1
 #define TCG_TARGET_HAS_sat_vec          1
 #define TCG_TARGET_HAS_minmax_vec       1
diff --git a/tcg/i386/tcg-target.h b/tcg/i386/tcg-target.h
index d2baf796b0..abd4ac7fc0 100644
--- a/tcg/i386/tcg-target.h
+++ b/tcg/i386/tcg-target.h
@@ -189,7 +189,6 @@ extern bool have_avx2;
 #define TCG_TARGET_HAS_shi_vec          1
 #define TCG_TARGET_HAS_shs_vec          1
 #define TCG_TARGET_HAS_shv_vec          have_avx2
-#define TCG_TARGET_HAS_cmp_vec          1
 #define TCG_TARGET_HAS_mul_vec          1
 #define TCG_TARGET_HAS_sat_vec          1
 #define TCG_TARGET_HAS_minmax_vec       1
diff --git a/tcg/ppc/tcg-target.h b/tcg/ppc/tcg-target.h
index aee38157a2..be10363956 100644
--- a/tcg/ppc/tcg-target.h
+++ b/tcg/ppc/tcg-target.h
@@ -169,7 +169,6 @@ extern bool have_vsx;
 #define TCG_TARGET_HAS_shi_vec          0
 #define TCG_TARGET_HAS_shs_vec          0
 #define TCG_TARGET_HAS_shv_vec          1
-#define TCG_TARGET_HAS_cmp_vec          1
 #define TCG_TARGET_HAS_mul_vec          1
 #define TCG_TARGET_HAS_sat_vec          1
 #define TCG_TARGET_HAS_minmax_vec       1
-- 
2.25.1



^ permalink raw reply related	[flat|nested] 14+ messages in thread

* [PULL 11/11] accel/tcg: Fix computing of is_write for MIPS
  2020-10-08 12:56 [PULL 00/11] tcg patch queue Richard Henderson
                   ` (9 preceding siblings ...)
  2020-10-08 12:56 ` [PULL 10/11] tcg: Remove TCG_TARGET_HAS_cmp_vec Richard Henderson
@ 2020-10-08 12:56 ` Richard Henderson
  2020-10-08 13:36 ` [PULL 00/11] tcg patch queue no-reply
  2020-10-08 18:13 ` Peter Maydell
  12 siblings, 0 replies; 14+ messages in thread
From: Richard Henderson @ 2020-10-08 12:56 UTC (permalink / raw)
  To: qemu-devel; +Cc: peter.maydell, Xu Zou, Kele Huang

From: Kele Huang <kele.hwang@gmail.com>

Detect all MIPS store instructions in cpu_signal_handler for all available
MIPS versions, and set is_write if encountering such store instructions.

This fixed the error while dealing with self-modified code for MIPS.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Kele Huang <kele.hwang@gmail.com>
Signed-off-by: Xu Zou <iwatchnima@gmail.com>
Message-Id: <20201002081420.10814-1-kele.hwang@gmail.com>
[rth: Use uintptr_t for pc to fix n32 build error.]
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/tcg/user-exec.c | 43 +++++++++++++++++++++++++++++++++++++++----
 1 file changed, 39 insertions(+), 4 deletions(-)

diff --git a/accel/tcg/user-exec.c b/accel/tcg/user-exec.c
index 5c96819ded..4ebe25461a 100644
--- a/accel/tcg/user-exec.c
+++ b/accel/tcg/user-exec.c
@@ -702,16 +702,51 @@ int cpu_signal_handler(int host_signum, void *pinfo,
 
 #elif defined(__mips__)
 
+#if defined(__misp16) || defined(__mips_micromips)
+#error "Unsupported encoding"
+#endif
+
 int cpu_signal_handler(int host_signum, void *pinfo,
                        void *puc)
 {
     siginfo_t *info = pinfo;
     ucontext_t *uc = puc;
-    greg_t pc = uc->uc_mcontext.pc;
-    int is_write;
+    uintptr_t pc = uc->uc_mcontext.pc;
+    uint32_t insn = *(uint32_t *)pc;
+    int is_write = 0;
+
+    /* Detect all store instructions at program counter. */
+    switch((insn >> 26) & 077) {
+    case 050: /* SB */
+    case 051: /* SH */
+    case 052: /* SWL */
+    case 053: /* SW */
+    case 054: /* SDL */
+    case 055: /* SDR */
+    case 056: /* SWR */
+    case 070: /* SC */
+    case 071: /* SWC1 */
+    case 074: /* SCD */
+    case 075: /* SDC1 */
+    case 077: /* SD */
+#if !defined(__mips_isa_rev) || __mips_isa_rev < 6
+    case 072: /* SWC2 */
+    case 076: /* SDC2 */
+#endif
+        is_write = 1;
+        break;
+    case 023: /* COP1X */
+        /* Required in all versions of MIPS64 since
+           MIPS64r1 and subsequent versions of MIPS32r2. */
+        switch (insn & 077) {
+        case 010: /* SWXC1 */
+        case 011: /* SDXC1 */
+        case 015: /* SUXC1 */
+            is_write = 1;
+        }
+        break;
+    }
 
-    /* XXX: compute is_write */
-    is_write = 0;
     return handle_cpu_signal(pc, info, is_write, &uc->uc_sigmask);
 }
 
-- 
2.25.1



^ permalink raw reply related	[flat|nested] 14+ messages in thread

* Re: [PULL 00/11] tcg patch queue
  2020-10-08 12:56 [PULL 00/11] tcg patch queue Richard Henderson
                   ` (10 preceding siblings ...)
  2020-10-08 12:56 ` [PULL 11/11] accel/tcg: Fix computing of is_write for MIPS Richard Henderson
@ 2020-10-08 13:36 ` no-reply
  2020-10-08 18:13 ` Peter Maydell
  12 siblings, 0 replies; 14+ messages in thread
From: no-reply @ 2020-10-08 13:36 UTC (permalink / raw)
  To: richard.henderson; +Cc: peter.maydell, qemu-devel

Patchew URL: https://patchew.org/QEMU/20201008125659.49857-1-richard.henderson@linaro.org/



Hi,

This series seems to have some coding style problems. See output below for
more information:

Type: series
Message-id: 20201008125659.49857-1-richard.henderson@linaro.org
Subject: [PULL 00/11] tcg patch queue

=== TEST SCRIPT BEGIN ===
#!/bin/bash
git rev-parse base > /dev/null || exit 0
git config --local diff.renamelimit 0
git config --local diff.renames True
git config --local diff.algorithm histogram
./scripts/checkpatch.pl --mailback base..
=== TEST SCRIPT END ===

Updating 3c8cf5a9c21ff8782164d1def7f44bd888713384
From https://github.com/patchew-project/qemu
 - [tag update]      patchew/20200930121105.667049-1-eblake@redhat.com -> patchew/20200930121105.667049-1-eblake@redhat.com
 - [tag update]      patchew/20201001170752.82063-1-richard.henderson@linaro.org -> patchew/20201001170752.82063-1-richard.henderson@linaro.org
 - [tag update]      patchew/20201002215955.254866-1-richard.henderson@linaro.org -> patchew/20201002215955.254866-1-richard.henderson@linaro.org
 - [tag update]      patchew/20201004180443.2035359-1-f4bug@amsat.org -> patchew/20201004180443.2035359-1-f4bug@amsat.org
 - [tag update]      patchew/20201007160038.26953-1-alex.bennee@linaro.org -> patchew/20201007160038.26953-1-alex.bennee@linaro.org
 * [new tag]         patchew/20201008125659.49857-1-richard.henderson@linaro.org -> patchew/20201008125659.49857-1-richard.henderson@linaro.org
Switched to a new branch 'test'
9aed6a6 accel/tcg: Fix computing of is_write for MIPS
4d1e622 tcg: Remove TCG_TARGET_HAS_cmp_vec
c4897cd tcg/optimize: Fold dup2_vec
2e58dc0 tcg: Fix generation of dupi_vec for 32-bit host
96f7941 tcg/i386: Fix dupi for avx2 32-bit hosts
0b5bcbb tcg: Remove TCGOpDef.used
40cea0e tcg: Move some TCG_CT_* bits to TCGArgConstraint bitfields
7c62277 tcg: Remove TCG_CT_REG
42d330b tcg: Move sorted_args into TCGArgConstraint.sort_index
1eff219 tcg: Drop union from TCGArgConstraint
2c84345 tcg: Adjust simd_desc size encoding

=== OUTPUT BEGIN ===
1/11 Checking commit 2c84345198c1 (tcg: Adjust simd_desc size encoding)
2/11 Checking commit 1eff219a2edb (tcg: Drop union from TCGArgConstraint)
ERROR: braces {} are necessary for all arms of this statement
#407: FILE: tcg/tcg.c:2213:
+            if (tcg_regset_test_reg(arg_ct->regs, i))
[...]

total: 1 errors, 0 warnings, 420 lines checked

Patch 2/11 has style problems, please review.  If any of these errors
are false positives report them to the maintainer, see
CHECKPATCH in MAINTAINERS.

3/11 Checking commit 42d330b3f312 (tcg: Move sorted_args into TCGArgConstraint.sort_index)
4/11 Checking commit 7c62277f9d05 (tcg: Remove TCG_CT_REG)
5/11 Checking commit 40cea0e3c162 (tcg: Move some TCG_CT_* bits to TCGArgConstraint bitfields)
ERROR: spaces prohibited around that ':' (ctx:WxW)
#36: FILE: include/tcg/tcg.h:982:
+    unsigned ct : 16;
                 ^

ERROR: spaces prohibited around that ':' (ctx:WxW)
#37: FILE: include/tcg/tcg.h:983:
+    unsigned alias_index : 4;
                          ^

ERROR: spaces prohibited around that ':' (ctx:WxW)
#38: FILE: include/tcg/tcg.h:984:
+    unsigned sort_index : 4;
                         ^

ERROR: spaces prohibited around that ':' (ctx:WxW)
#39: FILE: include/tcg/tcg.h:985:
+    bool oalias : 1;
                 ^

ERROR: spaces prohibited around that ':' (ctx:WxW)
#40: FILE: include/tcg/tcg.h:986:
+    bool ialias : 1;
                 ^

ERROR: spaces prohibited around that ':' (ctx:WxW)
#41: FILE: include/tcg/tcg.h:987:
+    bool newreg : 1;
                 ^

total: 6 errors, 0 warnings, 106 lines checked

Patch 5/11 has style problems, please review.  If any of these errors
are false positives report them to the maintainer, see
CHECKPATCH in MAINTAINERS.

6/11 Checking commit 0b5bcbbf91ca (tcg: Remove TCGOpDef.used)
7/11 Checking commit 96f79419c763 (tcg/i386: Fix dupi for avx2 32-bit hosts)
8/11 Checking commit 2e58dc08335c (tcg: Fix generation of dupi_vec for 32-bit host)
9/11 Checking commit c4897cd9e9b1 (tcg/optimize: Fold dup2_vec)
10/11 Checking commit 4d1e62279de6 (tcg: Remove TCG_TARGET_HAS_cmp_vec)
11/11 Checking commit 9aed6a69f26e (accel/tcg: Fix computing of is_write for MIPS)
WARNING: architecture specific defines should be avoided
#28: FILE: accel/tcg/user-exec.c:705:
+#if defined(__misp16) || defined(__mips_micromips)

ERROR: space required before the open parenthesis '('
#44: FILE: accel/tcg/user-exec.c:719:
+    switch((insn >> 26) & 077) {

WARNING: architecture specific defines should be avoided
#57: FILE: accel/tcg/user-exec.c:732:
+#if !defined(__mips_isa_rev) || __mips_isa_rev < 6

WARNING: Block comments use a leading /* on a separate line
#64: FILE: accel/tcg/user-exec.c:739:
+        /* Required in all versions of MIPS64 since

WARNING: Block comments use * on subsequent lines
#65: FILE: accel/tcg/user-exec.c:740:
+        /* Required in all versions of MIPS64 since
+           MIPS64r1 and subsequent versions of MIPS32r2. */

WARNING: Block comments use a trailing */ on a separate line
#65: FILE: accel/tcg/user-exec.c:740:
+           MIPS64r1 and subsequent versions of MIPS32r2. */

total: 1 errors, 5 warnings, 55 lines checked

Patch 11/11 has style problems, please review.  If any of these errors
are false positives report them to the maintainer, see
CHECKPATCH in MAINTAINERS.

=== OUTPUT END ===

Test command exited with code: 1


The full log is available at
http://patchew.org/logs/20201008125659.49857-1-richard.henderson@linaro.org/testing.checkpatch/?type=message.
---
Email generated automatically by Patchew [https://patchew.org/].
Please send your feedback to patchew-devel@redhat.com

^ permalink raw reply	[flat|nested] 14+ messages in thread

* Re: [PULL 00/11] tcg patch queue
  2020-10-08 12:56 [PULL 00/11] tcg patch queue Richard Henderson
                   ` (11 preceding siblings ...)
  2020-10-08 13:36 ` [PULL 00/11] tcg patch queue no-reply
@ 2020-10-08 18:13 ` Peter Maydell
  12 siblings, 0 replies; 14+ messages in thread
From: Peter Maydell @ 2020-10-08 18:13 UTC (permalink / raw)
  To: Richard Henderson; +Cc: QEMU Developers

On Thu, 8 Oct 2020 at 13:57, Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> The following changes since commit 6eeea6725a70e6fcb5abba0764496bdab07ddfb3:
>
>   Merge remote-tracking branch 'remotes/huth-gitlab/tags/pull-request-2020-10-06' into staging (2020-10-06 21:13:34 +0100)
>
> are available in the Git repository at:
>
>   https://github.com/rth7680/qemu.git tags/pull-tcg-20201008
>
> for you to fetch changes up to 62475e9d007d83db4d0a6ccebcda8914f392e9c9:
>
>   accel/tcg: Fix computing of is_write for MIPS (2020-10-08 05:57:32 -0500)
>
> ----------------------------------------------------------------
> Extend maximum gvec vector size
> Fix i386 avx2 dupi
> Fix mips host user-only write detection
> Misc cleanups.
>



Applied, thanks.

Please update the changelog at https://wiki.qemu.org/ChangeLog/5.2
for any user-visible changes.

-- PMM


^ permalink raw reply	[flat|nested] 14+ messages in thread

end of thread, other threads:[~2020-10-08 18:14 UTC | newest]

Thread overview: 14+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2020-10-08 12:56 [PULL 00/11] tcg patch queue Richard Henderson
2020-10-08 12:56 ` [PULL 01/11] tcg: Adjust simd_desc size encoding Richard Henderson
2020-10-08 12:56 ` [PULL 02/11] tcg: Drop union from TCGArgConstraint Richard Henderson
2020-10-08 12:56 ` [PULL 03/11] tcg: Move sorted_args into TCGArgConstraint.sort_index Richard Henderson
2020-10-08 12:56 ` [PULL 04/11] tcg: Remove TCG_CT_REG Richard Henderson
2020-10-08 12:56 ` [PULL 05/11] tcg: Move some TCG_CT_* bits to TCGArgConstraint bitfields Richard Henderson
2020-10-08 12:56 ` [PULL 06/11] tcg: Remove TCGOpDef.used Richard Henderson
2020-10-08 12:56 ` [PULL 07/11] tcg/i386: Fix dupi for avx2 32-bit hosts Richard Henderson
2020-10-08 12:56 ` [PULL 08/11] tcg: Fix generation of dupi_vec for 32-bit host Richard Henderson
2020-10-08 12:56 ` [PULL 09/11] tcg/optimize: Fold dup2_vec Richard Henderson
2020-10-08 12:56 ` [PULL 10/11] tcg: Remove TCG_TARGET_HAS_cmp_vec Richard Henderson
2020-10-08 12:56 ` [PULL 11/11] accel/tcg: Fix computing of is_write for MIPS Richard Henderson
2020-10-08 13:36 ` [PULL 00/11] tcg patch queue no-reply
2020-10-08 18:13 ` Peter Maydell

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).